> For the complete documentation index, see [llms.txt](https://docs.lleverage.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.lleverage.ai/run-operations/agent/agent-context/memory.md). # Memory Understand how persistent Memory can help the Agent across sessions, what the three layers mean and where authoritative business truth should live. Memory helps the Agent carry useful context across sessions instead of treating every conversation as completely new. It is designed for persistent working context. It is **not** a replacement for the ERP, a maintained policy, a governed mapping or another authoritative source. ## Memory can be created and retrieved automatically Supported Agent work can contribute persistent context after a session, including extracted facts and reusable reflections from the conversation. Relevant Memory can also be retrieved automatically for later Agent runs when it matches the work being done. You do not need to paste the entire history into every new session or explicitly ask the Agent to remember every useful detail. This process is selective. Memory is not a complete transcript, and the fact that something was said once does not make it a guaranteed durable rule. ## The three Memory layers The current Memory model separates persistent context into: * **Observations** — raw or recent context such as past sessions, reflections and ingested material. * **Memories** — distilled reusable facts, preferences and rules of thumb. * **Knowledge** — more deliberately curated reference context. The layers help distinguish how context was formed. They are not a universal ranking of truth. A current ERP value can still outrank a remembered value. An approved policy can outrank an observation. A customer-specific rule can outrank generic industry guidance. ## Memory and Skills play different roles When an explicit Skill is active, its task-specific method should guide how the recurring job is performed. Memory can still contribute useful context, but it should not silently override the reviewed method in the Skill. If a remembered rule repeatedly conflicts with a Skill or maintained source, correct the owning layer rather than relying on the Agent to reconcile the contradiction differently each time. ## What operators see in a session In a web Agent session, the **Memory** area in the right-hand rail gives visibility into persistent context around the current work. Use it as supporting evidence. It is not intended to be a perfect trace of every retrieval decision or a guarantee that everything shown was decisive to the final answer. If a result surprises you, inspect the session's Context and Memory and then verify against the maintained source that should own the fact. ## What builders can manage **Intelligence → Memory** is the broader Project surface for inspecting and maintaining persistent context. Builders can use that surface to inspect and correct context that has become stale, misleading or too broad. ## Put durable information in the right home A useful rule: * exact operational record → source system; * maintained structured mapping → Table; * approved reference material → Knowledge Base; * reusable operating method → Skill; * useful persistent Project or session context → Memory. If the same remembered fact is important enough that it must be exact, governed or systematically updated, move it to the maintained source that should own it. ## Memory and channels Channel conversations do not turn all memory into one unrestricted organisation-wide pool. The run still uses authorised scope and conversation context. Personal and shared conversations can have different memory boundaries, and Project-specific information remains subject to the user's access. When the work should clearly stay inside one Project, pin the Slack or Teams conversation to that Project. --- # Agent Instructions This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com. ## Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter: ``` GET https://docs.lleverage.ai/run-operations/agent/agent-context/memory.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.